AI and copyright: part I

Is there copyright on my AI-generated works?

By Ariane De Croo Senior Trademarks | Designs Attorney IP Hills | Ipsilon Group

We have been using artificial intelligence (AI) for some time to make tasks easier or more efficient: drafting a text and then translating it, generating a layout, summarising or visualising documents…

Every day, around 34 million images are created with AI. The use and evolution of AI continues to develop exponentially, day by day.

More and more often, we get the question from clients whether there is copyright on images generated with AI.

Is there copyright on my AI-generated works?

A work may enjoy copyright protection if the creation:

• belongs to the field of literature or art
• is original
• is expressed in a concrete form and
• is the result of a creative activity.

Central to copyright is human creation – especially from European case law, it appears that a work must be an “expression of the creator’s own intellectual creation“. Thereby, the Painer judgment further clarifies that an intellectual creation of its own is the expression of free and creative choice, bearing the stamp of the author’s personality.

Thus, the key question is whether, in the face of interaction with AI or Generative AI (GAI), the human author has sufficiently self-created the output.

We also prefer the word choice ‘creator’ rather than ‘author’ in the classical sense in this case.

Jurisprudence on the matter is still evolving for obvious reasons.

We already see that in the United States of America, the US Copyright Office (USCO) takes a rather strict and reserved position. A lack of human creative control and thus a weak input from the human creator with a strong AI contribution ensure that no copyright will apply.

In contrast, in Chinese jurisprudence, a GAI model is considered a passive tool – like a classic software or camera – so the creative control would indeed lie with the human creator. The more detailed and elaborate the prompt, the more control and say over the output by the creator. Extensive and complex creative choices made by human creators during the AI-assisted ‘image generation process’ may thus meet the requirements of intellectual performance and originality for copyright protection.

However, we need to nuance this picture. The results of the prompt can be infinitely variable and unpredictable, even for highly elaborate prompts.

Adjustments on the output by the human creator that add sufficient personal stamp give more weight to the rights of the ‘author’.

For this reason, an ‘editorial’ phase through creative editing or post-editing by the human creator increases the likelihood that an AI image will fall under copyright protection.

How do we evaluate the collaboration or interaction between the creator and the machine (AI)?

Three options present themselves.

As long as AI is used purely as a tool, there will be little debate. The expression of the creator’s personality is clearly demonstrable. We can make the comparison with software used to create designs, for example.

It becomes different when AI is seen as a ‘producer’, especially when the outcome (output) is purely based on a prompt. These are quasi-autonomous AI productions. The creator provides weak or insufficient input, a personal stamp of the creator on the output is missing. Where appropriate, there will be no copyright protection for the output.

Less obvious is the third option where an interaction or dialogue arises between machine and creator: software written or developed by a human creator, data entered and selected by a human creator and language models trained by human creators…. The line is very difficult to draw between human input and AI results in output.

Human creation is central to protection under copyright law. For assessment, the legal criterion of originality will be applied: if free and creative choices are found, they provide an individual expression of the author and a link to the author’s personality. Of importance, therefore, are the personality of the creator and its demonstrability either in the process or in the output.

Author’s opinion

The rapidly evolving nature of AI technology means that legal frameworks remain in constant flux. Globally, the judiciary and legislators are working on a coherent approach to these new challenges.

It is expected that in the coming years, the European judiciary will strive to strike a balance between a strict test of human creativity and the reality of technological progress. In practice, the degree of human interference will be particularly decisive. The more clearly the creator’s creative choices and reworkings can be demonstrated, the greater the likelihood of protection. AI itself will in all likelihood not be recognised as an ‘author’, but it does constitute a powerful tool within a protected creative process.

These legal developments tie in with broader questions about the nature of creativity, authorship and the role of technology in human expression. That discussion will undoubtedly continue to evolve as AI continues to develop and become more deeply interwoven into creative workflows.

To be continued…

Sources:

AI Image Statistics for 2024: How Much Content Was Created by AI

ECJ EU 16 July 2009, C-5/08, ECLI:EU:C:2009:465 (Infopaq I).

ECJ EU 1 December 2011, C-145/10, ECLI:EU:C:2011:798, (Painer).

Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, US Government Information, Federal Register Vol. 88, No, 51, 16 March 2023, p. 16192.

Beijing Internet Court, 27 November 2023, Jing 0491 Min Chu No. 11279. See further: China’s Controversial Court Rulings on AI Output-and How It May Affect People in the US – Authors Alliance